3
$\begingroup$

I am trying to produce ideal anisotropic diffusion outputs using the standard lena image. Using the first coefficient , the output is coming somewhat close to ideal ouput but when I increase the kappa value the bluring increases. When I am using the coefficient that prefers wide area diffusion over small areas,my output comes blurred dominantly. I am providing the output images.!

INPUT: enter image description here

DESIRED OUTPUT: enter image description here

OUTPUT1:

output1

OUTPUT2:

out2

GOT A CLOSE OUTPUT: I got this output with kappa 10 ,iterations 80 and selecting the wide area coefficient.

How do i remove this blurring in images? because its hindering my research

$\endgroup$
  • $\begingroup$ Hi Aviral. I am not sure how to answer your question. What is your desired goal? What is the image you start with, and what is the image you want to reach? As for controlling the output in AD: the numerical value of Kappa controls the gradient threshold: that is the "barrier" above which diffusion will not occur. See it as a difference between pixel value in 2 zones. I guess that in the 2nd case, you set it to a too high value. $\endgroup$ – Jean-Yves Feb 5 '12 at 10:15
  • $\begingroup$ I edited the question Jean. The desired output is with kappa value=10 for 80 iterations. $\endgroup$ – Aviral Kumar Feb 5 '12 at 10:55
  • $\begingroup$ Ok here is what I would do. I will not make a proper answer for it, because that's something I would experiment with. The number one feature I see in your source image is a very strong *salt & pepper& noise on top of the other noise, and you retrieve some white and black pixels in your output image. This means that you had to use a strong kappa value to get rid of it, over-blurring edges. I do not think you cannot escape with AD alone. So before you try another filter, I would make a first step of MEDIAN FILTERING before AD $\endgroup$ – Jean-Yves Feb 5 '12 at 12:07
  • $\begingroup$ Hey Jean, the initial image dint had any noise. I added some salt and pepper noise. I applied median filtering too, but got little success. Check my latest edit. I got the close output with the wide area coefficent (kappa=10,iterations=80). $\endgroup$ – Aviral Kumar Feb 5 '12 at 13:13
  • $\begingroup$ sorry, that's as far as I can go with this. Btw: It's Jean-Yves. $\endgroup$ – Jean-Yves Feb 6 '12 at 13:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.